### Table 1: PreserveRt violations in losing and winning candidates for /hadinu+red/

2003

"... In PAGE 12: ... Thus the analysis uses a constraint PreserveRt/Lft which is implementable as a nite state machine and usurps the functions of both Contiguity and Anchor. PreserveRt outputs a violation for each MaxBR vio- lation after the rst non-violation of MaxBR within the reduplicant, and similarly PreserveLft outputs a violation for each MaxBR violation before the rightmost non-violation of MaxBR (See Table1 ) . Since PreserveLft/Rt can be satis ed by an empty reduplicant, the analysis also uses a simple existential faithfulness constraint to ensure that the reduplicant is not empty (this constraint, ExtMaxIR, is violated if there is any segment in the underlying reduplicant but no segments in the surface reduplicant note that ExtMaxBR would work just as well for this) (cf.... ..."

Cited by 3

### Tableau 2 In cases of more than one secondary stress, the constraint PK-MAX (B/O) can show its force, once ALIGN (PRWD, L, FT, L) is satisfied by the first secondary stress. Therefore candidate (a) wins, where the stress peak of the stem on the fourth syllable is preserved, although candidate (b), with stress on the third, heavy syllable, would satisfy the WSP. On the other hand, we do not have stress preservation when ALIGN (PRWD, L, FT, L) is challenged:

### Tableau 1 Candidate (a) has a trochaic ( apos;LH) foot aligned with the left edge of the prosodic word. Therefore it wins against candidate (b), which would satisfy the WSP by stressing the second, heavy syllable. There are other candidates, which would satisfy both ALIGN (PRWD, L, FT, L) and the WSP, but which would violate constraints that turn out to be higher ranked than ALIGN (PRWD, L, FT, L).23 This is the case for the following parsing:

### Table 17. Percent of accurate search results.

in OF

"... In PAGE 7: ...able 16. Web prediction acuracies...................................46 Table17 .... In PAGE 55: ... For each of these deciles, the thre candidates whose number of search results was closest to the average of each of the deciles were examined. Table17 displays the name of the candidate, the total number of web search results, the number of results that correctly expresed the opinion that the candidate would win, the number of results examined, and the percentage that correctly expresed the opinion that the candidate would win. ... ..."

### Table 1: A simulation of ten elections where every voter votes for candidate 1 and 100 receipts are collected.

"... In PAGE 2: ... To show the efiect we wrote a small simulation program. Table1 shows ten simulations for an election with three candidates, where 100 receipts have been collected and candidate 1 gets all the votes. The lines show the number of marks for each candidate, leaving no doubt at all about who is winning already while voting is still going on.... ..."

### Table 1: A simulation of ten elections where every voter votes for candidate 1 and 100 receipts are collected.

"... In PAGE 2: ... To show the efiect we wrote a small simulation program. Table1 shows ten simulations for an election with three candidates, where 100 receipts have been collected and candidate 1 gets all the votes. The lines show the number of marks for each candidate, leaving no doubt at all about who is winning already while voting is still going on.... ..."

### Table 1 Results showing convergence of the \maximizing set quot; Sk (which contains the global maximizer of R[G(j!)]) for Example 1. Due to the symmetry, only quantities to the right of the origin are shown.

"... In PAGE 8: ... For this example, it turns out that Rk and Sk are intervals at every iteration. We show various quantities of interest in Table1 . As can be seen there, the cubic t candidate (for !k+1) wins over the midpoint candidate near the optimum, and the algorithm appears to be quadratically convergent.... ..."